Guides
GDP first print vs revisions — why the number changes
US GDP is not published once. The Bureau of Economic Analysis releases a sequence of estimates for each quarter — an advance estimate, followed by second and third estimates roughly a month apart — as more complete source data arrives. Annual and comprehensive updates then revise the series again, sometimes years later.
The first print is therefore an estimate built from incomplete inputs, and it is also the number the market trades. When the advance estimate lands, positions move against that figure, not against the value it will eventually settle at. Any study of how markets responded has to use the print that existed at the time.
A concrete illustration from the site's own example: for 2023 Q4, BEA's advance estimate of real GDP growth was 3.3 percent annualized, and the third estimate published two months later was 3.4 percent. A small revision, and enough to change the sign of a surprise measure built against a consensus that sat between them.
Vintages also matter well beyond the initial sequence. Benchmark and comprehensive revisions can change levels and growth rates across the whole history at once, which means a series downloaded today may differ from the same series downloaded last year for dates decades in the past. Only an archive that kept each release can tell you which version a decision was made against.
What's covered
- Advance, second and third estimates released roughly a month apart
- Annual and comprehensive updates revising the series afterwards
- The first print is what markets trade and what surprise is measured against
- Benchmark revisions can restate history across the entire series
- Vintage archives are the only way to recover an earlier release
- The same pattern applies to inflation, employment and activity data
Questions
Which vintage should I use for an event study?
The one that was current at the event. For a reaction to a release, that is the release itself; for anything measured in the days after, it is whatever had been published by then and nothing later.
Are revisions predictable?
Not reliably enough to lean on, and assuming a direction is its own bias. The safer discipline is to treat each vintage as the fact that existed at its time rather than as a noisy estimate of the final value.
Where can I check historical vintages myself?
ALFRED, the St. Louis Fed's archival companion to FRED, publishes vintage series for many US indicators, and BEA documents its own release and revision schedule. Both are linked below.
Sources
Related
Join the waitlist for early access.
Join the waitlist. Vote the roadmap. First in, first served.